WEBVTT - Google's Liz Reid Talks AI Mode in Search

0:00:02.759 --> 0:00:08.880
<v Speaker 1>Bloomberg Audio Studios, Podcasts, Radio News. All right, back to

0:00:09.000 --> 0:00:12.040
<v Speaker 1>kind of our tech focus this hour, our team around

0:00:12.080 --> 0:00:14.560
<v Speaker 1>the globe and across the United States, continuing to report

0:00:14.600 --> 0:00:16.440
<v Speaker 1>out on some of the big tech events that are

0:00:16.440 --> 0:00:18.800
<v Speaker 1>happening on this Tuesday. Right now, we want to head

0:00:18.840 --> 0:00:22.600
<v Speaker 1>to Google Io. It's Google's annual developer conference. It focuses

0:00:22.640 --> 0:00:26.520
<v Speaker 1>on the company's advancements in AI, Android operating system, smartwatches,

0:00:26.840 --> 0:00:29.920
<v Speaker 1>and it's scam detection tools. And Tim, it's happening right

0:00:29.920 --> 0:00:31.280
<v Speaker 1>now in Mountain View, California.

0:00:31.440 --> 0:00:34.480
<v Speaker 2>That's where we find Bloomberg News technology reporter Jack Devallos

0:00:34.560 --> 0:00:35.760
<v Speaker 2>with Google's Liz Reach.

0:00:35.800 --> 0:00:38.320
<v Speaker 1>He's vice president and head of Search at the firm.

0:00:38.360 --> 0:00:40.080
<v Speaker 1>Hey Jackie, Hey Tim.

0:00:39.920 --> 0:00:42.360
<v Speaker 2>And Carol Liz, it's great to have you today.

0:00:42.479 --> 0:00:44.040
<v Speaker 3>A pleasure to be with you here today.

0:00:44.280 --> 0:00:46.560
<v Speaker 2>The big news that came out of that keynote speech

0:00:46.680 --> 0:00:50.159
<v Speaker 2>was really AI Mode. It's what everyone was really hoping

0:00:50.520 --> 0:00:53.320
<v Speaker 2>to get an insight into what's next for search, and

0:00:53.440 --> 0:00:58.120
<v Speaker 2>you called it the most powerful search and Geniette described

0:00:58.200 --> 0:01:01.400
<v Speaker 2>to us why this is a total reimagining of the

0:01:01.400 --> 0:01:02.280
<v Speaker 2>search experience.

0:01:03.120 --> 0:01:06.600
<v Speaker 3>Yeah, I think we've seen with aio reviews that it's

0:01:06.600 --> 0:01:09.280
<v Speaker 3>been really great for people, enabling them to ask new

0:01:09.319 --> 0:01:12.119
<v Speaker 3>types of questions. But we also saw that people really

0:01:12.160 --> 0:01:14.400
<v Speaker 3>wanted to say sometimes they have those hard questions, they

0:01:14.400 --> 0:01:17.480
<v Speaker 3>want to go directly to an AI powered response. They

0:01:17.480 --> 0:01:20.039
<v Speaker 3>wanted to ask follow up questions because they're trying to

0:01:20.040 --> 0:01:22.160
<v Speaker 3>do something in more depth in research, and so we

0:01:22.240 --> 0:01:25.080
<v Speaker 3>really thought about what does that experience look like for people,

0:01:25.560 --> 0:01:27.080
<v Speaker 3>and how do we bring it in so that they

0:01:27.080 --> 0:01:30.200
<v Speaker 3>can ask truly their hardest questions and then continue to

0:01:30.200 --> 0:01:31.920
<v Speaker 3>explore and take action throughout the web.

0:01:32.200 --> 0:01:34.200
<v Speaker 2>So is this the end of the blue links?

0:01:34.800 --> 0:01:38.880
<v Speaker 3>No, absolutely not. I think both within aiover reviews and

0:01:38.920 --> 0:01:42.240
<v Speaker 3>within AI mode, AI is actually allowing us to go

0:01:42.360 --> 0:01:46.600
<v Speaker 3>deeper into search. It's allowing people to find that hyper relevant,

0:01:46.800 --> 0:01:50.400
<v Speaker 3>incredibly fresh content that really matches your need. If you

0:01:50.440 --> 0:01:52.520
<v Speaker 3>think about it, if you are doing more of sort

0:01:52.560 --> 0:01:54.840
<v Speaker 3>of a keywordy style, you only gave a few words.

0:01:54.880 --> 0:01:58.040
<v Speaker 3>You didn't really describe what the particular type of shoes

0:01:58.080 --> 0:02:01.559
<v Speaker 3>you want are. And you saw with the shopping example today, Okay,

0:02:01.560 --> 0:02:03.920
<v Speaker 3>if you're really describing this is the type of rug

0:02:03.960 --> 0:02:06.640
<v Speaker 3>I want to buy, can you find that great small

0:02:06.680 --> 0:02:09.240
<v Speaker 3>merchant that sells the perfect rug for you, and so

0:02:09.320 --> 0:02:11.360
<v Speaker 3>I think that will continue to thrive going forward.

0:02:11.600 --> 0:02:14.040
<v Speaker 2>One of the biggest questions is really around search has

0:02:14.080 --> 0:02:16.280
<v Speaker 2>really come up around do I use one of these

0:02:16.320 --> 0:02:21.399
<v Speaker 2>other chatbots open ai for example, or Meta whatever's out

0:02:21.400 --> 0:02:24.320
<v Speaker 2>there these days, to kind of ask it these questions

0:02:24.320 --> 0:02:26.760
<v Speaker 2>in the same way that I used to use Google

0:02:26.840 --> 0:02:29.960
<v Speaker 2>Search for when it comes to Google's products, why do

0:02:30.040 --> 0:02:33.080
<v Speaker 2>you want users? Do you prefer to have users go

0:02:33.200 --> 0:02:35.919
<v Speaker 2>through the Gemini app or to search? Kind of what's

0:02:35.960 --> 0:02:39.240
<v Speaker 2>the unique benefit of going to search in the case

0:02:39.280 --> 0:02:43.320
<v Speaker 2>for more specific and more complex questions. I think search

0:02:43.440 --> 0:02:46.560
<v Speaker 2>is really designed around helping you with your information needs.

0:02:47.360 --> 0:02:52.000
<v Speaker 2>It's tuned to really think about things like accuracy and factuality. Oftentimes,

0:02:52.000 --> 0:02:54.320
<v Speaker 2>when you have these information needs, you want to get started,

0:02:54.320 --> 0:02:56.320
<v Speaker 2>but then you want to go deeper and research more.

0:02:57.000 --> 0:02:59.120
<v Speaker 3>Let's say you're looking for camps. You want some good

0:02:59.120 --> 0:03:00.840
<v Speaker 3>idea for camps, but we actually want to check out

0:03:00.880 --> 0:03:04.000
<v Speaker 3>the camp sitans, right, and so search is really designed

0:03:04.080 --> 0:03:07.359
<v Speaker 3>to make those questions possible and really focused on doing

0:03:07.360 --> 0:03:10.079
<v Speaker 3>the best job possible when you have questions with really

0:03:10.120 --> 0:03:11.160
<v Speaker 3>at information focus.

0:03:11.680 --> 0:03:16.000
<v Speaker 2>You mentioned accuracy, and Google had some issues with accuracy

0:03:16.160 --> 0:03:19.520
<v Speaker 2>kind of in the more early days of this technology.

0:03:19.680 --> 0:03:22.240
<v Speaker 2>How did you know that AI mode was ready for

0:03:22.320 --> 0:03:22.840
<v Speaker 2>prime time?

0:03:23.840 --> 0:03:27.080
<v Speaker 3>So we have quite thorough testing that we do constantly.

0:03:27.560 --> 0:03:31.040
<v Speaker 3>We have the feedback from the users in the Labs product,

0:03:31.560 --> 0:03:35.200
<v Speaker 3>but also put it through extensive testing around the factuality,

0:03:35.880 --> 0:03:38.520
<v Speaker 3>around the helpfulness of the product, and really seeing that

0:03:38.560 --> 0:03:41.480
<v Speaker 3>both from our benchmarks as well as from the user feedback,

0:03:41.520 --> 0:03:42.880
<v Speaker 3>we believe it is ready.

0:03:43.760 --> 0:03:47.720
<v Speaker 2>The agent kind of aspect to artificial intelligence is something

0:03:47.800 --> 0:03:51.480
<v Speaker 2>that everyone is really curious about when it comes to search.

0:03:52.000 --> 0:03:55.120
<v Speaker 2>What do agents mean in that context? Is it going

0:03:55.160 --> 0:03:58.880
<v Speaker 2>to be kind of us feeding information into search or

0:03:59.000 --> 0:04:01.440
<v Speaker 2>kind of it doing it for us in the long run.

0:04:02.600 --> 0:04:05.000
<v Speaker 3>I think maybe a couple of things to highlight, and

0:04:05.040 --> 0:04:08.040
<v Speaker 3>we talked about how we're bringing Project Mariner in today

0:04:08.080 --> 0:04:10.800
<v Speaker 3>to search. On those cases. For instance, you're trying to

0:04:10.800 --> 0:04:13.640
<v Speaker 3>buy baseball tickets, you would have to go to hundreds

0:04:13.680 --> 0:04:16.560
<v Speaker 3>of different sites enter in the same day the same

0:04:16.640 --> 0:04:18.840
<v Speaker 3>number of tickets and go look at them, and we

0:04:18.880 --> 0:04:21.360
<v Speaker 3>can go off and help pull that information together and

0:04:21.400 --> 0:04:23.960
<v Speaker 3>then allow you to take action. But I think this

0:04:24.080 --> 0:04:26.560
<v Speaker 3>is an important part of agents is agents should be

0:04:26.680 --> 0:04:29.280
<v Speaker 3>taking the drudgery out of the work for you. That's

0:04:29.320 --> 0:04:31.200
<v Speaker 3>how we think about in search, but still keeping you

0:04:31.240 --> 0:04:33.760
<v Speaker 3>in control. Right, So what's the part of the task

0:04:33.839 --> 0:04:36.000
<v Speaker 3>you don't want to do, but then coming back and saying, well,

0:04:36.000 --> 0:04:38.520
<v Speaker 3>which of these do you want? Right? I think if

0:04:38.520 --> 0:04:41.080
<v Speaker 3>you're trying to pick tickets or you're trying to buy

0:04:41.400 --> 0:04:44.320
<v Speaker 3>a great purse, both of those, at the end of

0:04:44.360 --> 0:04:47.359
<v Speaker 3>the day, you probably still want to sign off on

0:04:47.400 --> 0:04:49.880
<v Speaker 3>that final choice. And so we're really thinking about the

0:04:50.000 --> 0:04:53.159
<v Speaker 3>role of both reducing the effort but enabling you the

0:04:53.240 --> 0:04:54.320
<v Speaker 3>agency to participate.

0:04:54.560 --> 0:04:57.719
<v Speaker 2>I'm hearing a lot about engagement, so users are going

0:04:57.800 --> 0:04:59.720
<v Speaker 2>to be using this in a different way, possibly for

0:04:59.800 --> 0:05:03.680
<v Speaker 2>law longer and more efficiently. Earlier this year, an Apple

0:05:03.760 --> 0:05:06.039
<v Speaker 2>executives said that the number of Google searches in the

0:05:06.080 --> 0:05:09.320
<v Speaker 2>Safari browser was starting to take down. Do you expect

0:05:09.360 --> 0:05:11.760
<v Speaker 2>that AI mode could actually reverse that trend?

0:05:12.800 --> 0:05:16.520
<v Speaker 3>So, as we stated, overall search queries are up, including

0:05:16.560 --> 0:05:19.640
<v Speaker 3>on the Apple platforms. But I think you also heard

0:05:19.640 --> 0:05:23.719
<v Speaker 3>from Sender today aioverviews has driven an increase in queries

0:05:23.720 --> 0:05:26.520
<v Speaker 3>already today. I think we expect with AI mode that

0:05:26.640 --> 0:05:30.080
<v Speaker 3>will happen as well. It's really an expansionary moment. There's

0:05:30.120 --> 0:05:32.120
<v Speaker 3>a lot of questions that you might think about in

0:05:32.160 --> 0:05:34.440
<v Speaker 3>your head, but you didn't historically think it was worth

0:05:34.440 --> 0:05:37.400
<v Speaker 3>the time or effort to go ask them. And as

0:05:37.440 --> 0:05:40.640
<v Speaker 3>technology unlocks this, we saw this with aio reviews. People

0:05:40.720 --> 0:05:42.760
<v Speaker 3>just come and they asked that question that used to

0:05:42.800 --> 0:05:44.440
<v Speaker 3>pop in their head, and they would let go and

0:05:44.480 --> 0:05:46.200
<v Speaker 3>do that. And so we really expect the space to

0:05:46.200 --> 0:05:46.880
<v Speaker 3>grow over time.

0:05:47.200 --> 0:05:50.120
<v Speaker 2>One of the areas that I know publishers and content

0:05:50.120 --> 0:05:52.640
<v Speaker 2>creators are really curious about is kind of what the

0:05:52.720 --> 0:05:57.039
<v Speaker 2>longer term impact of aioverviews is on them. Can you

0:05:57.120 --> 0:06:00.600
<v Speaker 2>kind of break down how you expect AI mode kind

0:06:00.600 --> 0:06:03.400
<v Speaker 2>of largely available to affect that cohort.

0:06:04.400 --> 0:06:07.200
<v Speaker 3>So I would say two things we saw with aioverviews

0:06:08.560 --> 0:06:12.080
<v Speaker 3>that pages with aioverviews drive higher quality clicks to websites,

0:06:12.400 --> 0:06:14.400
<v Speaker 3>which means people spend more time on them, they're really

0:06:14.400 --> 0:06:16.280
<v Speaker 3>engaged on them, So it's a good thing for publishers.

0:06:16.279 --> 0:06:18.400
<v Speaker 3>So I think that's a great thing of getting really

0:06:18.480 --> 0:06:21.000
<v Speaker 3>high quality content, and we are looking forward to doing

0:06:21.040 --> 0:06:23.840
<v Speaker 3>the same thing with AI mode as well. I think

0:06:23.880 --> 0:06:26.640
<v Speaker 3>we're going to continue to find deeper parts of the

0:06:26.680 --> 0:06:30.200
<v Speaker 3>web that really incredible, hyper relevant content. But I would

0:06:30.240 --> 0:06:32.800
<v Speaker 3>also say I still expect the vast majority of people

0:06:33.520 --> 0:06:37.120
<v Speaker 3>to interact primarily on the main search experience with aioverviews.

0:06:37.360 --> 0:06:40.880
<v Speaker 3>AI mode is really for those harder tasks. It's for

0:06:41.120 --> 0:06:42.960
<v Speaker 3>the subset of people that really want to be on

0:06:43.000 --> 0:06:46.160
<v Speaker 3>that sort of cutting edge frontier of models. But AIO

0:06:46.240 --> 0:06:48.359
<v Speaker 3>reviews and the main search page continues to be a

0:06:48.360 --> 0:06:51.440
<v Speaker 3>great place for a lot of your questions. When it

0:06:51.440 --> 0:06:55.160
<v Speaker 3>comes to antitrust, it's on our viewer's minds. We're Bloomberg

0:06:55.240 --> 0:06:57.440
<v Speaker 3>and we're really focused on what those implications could be

0:06:57.520 --> 0:07:01.400
<v Speaker 3>on the broader business model. Judge last year ruled that

0:07:01.520 --> 0:07:04.920
<v Speaker 3>Google's dominance in search was an illegal monopoly, and of

0:07:04.960 --> 0:07:08.640
<v Speaker 3>course the outcome of that is still really yet to

0:07:08.640 --> 0:07:11.400
<v Speaker 3>be seen. But for you in your shoes as head

0:07:11.400 --> 0:07:13.640
<v Speaker 3>of search, how are you thinking about the effect that

0:07:13.680 --> 0:07:14.240
<v Speaker 3>could have.

0:07:14.600 --> 0:07:15.800
<v Speaker 2>On your roadmap?

0:07:16.680 --> 0:07:19.240
<v Speaker 3>So in general, we stay just focused on building the

0:07:19.240 --> 0:07:22.520
<v Speaker 3>best thing for our users. People continue to choose Google

0:07:22.720 --> 0:07:24.880
<v Speaker 3>because they find it helpful, not because they have to,

0:07:25.480 --> 0:07:27.560
<v Speaker 3>and that is really our north star and has been

0:07:27.640 --> 0:07:30.000
<v Speaker 3>for years. You go back to that mission of organizing

0:07:30.040 --> 0:07:33.560
<v Speaker 3>all the world's information and make it universally accessible and useful.

0:07:34.120 --> 0:07:38.440
<v Speaker 3>We're a long ways away from the end of that right, Yes,

0:07:38.560 --> 0:07:40.520
<v Speaker 3>we've come a long way in the last twenty five years,

0:07:40.520 --> 0:07:43.640
<v Speaker 3>but there's still so many opportunities to make information more useful.

0:07:44.320 --> 0:07:46.280
<v Speaker 3>We talk today about how we think the future of

0:07:46.280 --> 0:07:49.880
<v Speaker 3>search is going beyond information to intelligence, really helping you

0:07:49.960 --> 0:07:52.000
<v Speaker 3>in a deep way with your needs, and that's that's

0:07:52.000 --> 0:07:53.600
<v Speaker 3>what gets us motivated every day.

0:07:53.760 --> 0:07:56.880
<v Speaker 2>So you don't see the outcome of that antitrust trial

0:07:57.000 --> 0:08:01.520
<v Speaker 2>kind of affecting your strategy in search going forward or

0:08:01.640 --> 0:08:02.720
<v Speaker 2>slowing it down in any way.

0:08:03.040 --> 0:08:06.880
<v Speaker 3>So we don't generally comment on pending legislation, but I

0:08:06.920 --> 0:08:10.280
<v Speaker 3>think in general we'll continue to try and build amazing

0:08:10.320 --> 0:08:12.600
<v Speaker 3>products for users that people love, just as we have

0:08:12.680 --> 0:08:14.120
<v Speaker 3>over the last twenty five years.

0:08:14.280 --> 0:08:15.760
<v Speaker 2>Liz Rae, thanks so much for joining us.

0:08:15.800 --> 0:08:17.400
<v Speaker 3>Thank you for joining having me Carol